DIPARTIMENTO DI MEDICINA E CHIRURGIA
CORSO DI LAUREA MAGISTRALE IN PSICOBIOLOGIA E NEUROSCIENZE COGNITIVE
F ROM M IRROR N EURONS TO R EHABILITATION: E FFECTS OF A PPROACHES B ASED ON A CTION O BSERVATION
D AI N EURONI S PECCHIO ALLA R IABILITAZIONE: E FFETTI DEGLI A PPROCCI B ASATI SULL’ O SSERVAZIONE DI A ZIONI
Relatore:
Chiar.mo Prof. LUCA BONINI
Controrelatore:
Chiar.ma Prof.ssa ANNALISA PELOSI
Laureanda:
CRISTINA ROTUNNO
ANNO ACCADEMICO 2018/2019
2
3 INDEX
Summary ... 5
1. Introduction ... 7
1.1 The discovery of mirror neurons in the monkey ... 8
1.1.1 Basic properties of mirror neurons in the monkey ... 10
1.2 Evidence of a “mirror neuron system” in the human brain ... 12
1.3 Evidence for view-dependent coding of observed actions ... 16
1.4 Interaction between viewpoint and spatial location of observed action ... 18
1.5 Application of neuroscientific findings to human neurorehabilitation ... 19
1.5.1 Rehabilitation approaches: vicariate or restoring motor functions? ... 20
1.5.2 The Action Observation Treatment (AOT) ... 23
2. Objectives ... 26
3. Methods ... 27
4. Results ... 29
4.1 AOT for upper limb function in cerebral palsy ... 29
4.1.1 Identification of included studies ... 29
4.1.2 Characteristics of included studies ... 30
4.1.3 Action observation training and control conditions ... 30
4.1.4 Meta-analysis of AOT for upper limb function ... 31
4.2 AOT in stroke patients ... 32
4.2.1 Identification of included studies ... 32
4
4.2.2 Characteristics of included studies ... 33
4.2.3 Action observation training and control conditions ... 34
4.2.4 Meta-analysis of AOT for arm and hand function, walking ability and activities of daily living ... 34
4.3 AOT in Parkinson’s disease patients ... 36
4.4 AOT in orthopedic patients ... 38
5. Discussion ... 40
5.1 Control group ... 42
5.2 Videoclips ... 42
5.3 Viewpoint of observed action ... 43
5.4 Observed model... 44
References: ... 46
5 Summary
The discovery of mirror neurons in the monkey and of a mirror system in humans brought to the development of new rehabilitation approaches that leverage the property of these neurons and brain areas to become active both during the execution of goal-directed actions and the observation of similar actions performed by others. The aim of this work is to provide a comprehensive view on current findings derived from therapies based on action observation, and to provide information about their application in different groups of patients. Many studies investigating the efficacy of action observation treatment have been published, and its validity has been assessed in several neurological and non- neurological diseases. A quantitative analysis of studies carried out in the field of recovery of upper limb motor function in children with cerebral palsy with a meta-analytic approach has revealed a moderate functional improvement, consistent with the results obtained in other subpopulations of patients. These findings suggest that this approach may be successfully used to improve and enhance plastic processes in the motor system to promote motor recovery from different sources of damage.
6
La scoperta dei neuroni specchio nella scimmia di un sistema specchio nell’uomo ha portato allo sviluppo di nuove tecniche di riabilitazione che sfruttano la proprietà di questi neuroni ed aree cerebrali di attivarsi sia durante l’esecuzione di azioni finalizzate che durante l’osservazione di azioni simili eseguite da altri individui. Lo scopo di questo elaborato è quello di fornire una visione globale circa le attuali conoscenze sulle terapie basate sull’osservazione di azioni e fornire informazioni circa le loro attuali applicazioni in diversi gruppi di pazienti. Sono stati pubblicati molti studi che hanno indagato l’efficacia dell’action observation treatment e la sua validità è stata dimostrata nella riabilitazione sia di pazienti neurologici che non neurologici. Un’analisi quantitativa degli studi eseguiti nel campo del recupero della funzione motoria dell’arto superiore nei bambini con paralisi cerebrale infantile condotta con approccio meta-analitico ha rivelato un miglioramento funzionale di moderata entità, in accordo con i risultati ottenuti in altre sottopopolazioni di pazienti. Questi risultati suggeriscono che tale approccio può essere utilizzato con risultati promettenti per migliorare ed aumentare i processi di plasticità del sistema motorio per favorire il recupero motorio a partire da differenti tipi di danno.
7 1. Introduction
Since the discovery of mirror neurons (MNs), a class of visuomotor neurons that discharge both during the execution of specific goal-directed motor acts (e.g. grasping an object) and during the observation of another individual doing the same or a similar act (di Pellegrino et al., 1992; Gallese et al., 1996; Rizzolatti et al., 1996), the general view on the role of the motor system has radically changed. Indeed, beyond motor control, parieto-frontal circuits are nowadays thought to be involved in a multiplicity of high level perceptual and cognitive functions, such as object perception, space coding, action recognition, imitation, motor learning and interindividual communication.
Early studies on MNs in monkeys focused on their basic discharging properties, to detect the most appropriate stimuli to trigger a response. Subsequent studies found that their activation can be modulated by different factors, such as location in space or viewpoint of observed action.
The use of different techniques provided indirect evidence of the existence of a similar mirror mechanism also in the human brain. These findings had a profound impact in rehabilitation practice, through the idea that action observation, by the visual activation of motor representations, could facilitate functional recovery in patients with different kinds of motor impairments. This approach has been successfully applied to patients with stroke, patients with Parkinson’s disease, children with cerebral palsy and aphasic patients, but also to patients with non-neurological diseases, like those undergoing rehabilitation after an orthopedic surgery.
This work attempts to give a global view about the current knowledge of mirror mechanism in the human brain and its application in the neurorehabilitative field.
8
1.1 The discovery of mirror neurons in the monkey
Mirror neurons (MNs) were firstly discovered in the ventral premotor area F5 of the macaque (Figure 1). In 1992, di Pellegrino and coworkers described a class of visuo- motor cells in the monkey that were activated both during monkey’s goal-directed actions and during the observation of the same, or similar, experimenter’s actions, in the absence of any overt movement of the animal (di Pellegrino et al., 1992). Four years later, a paper was published (Gallese et al., 1996) describing extensively the properties of this class of cells. According to the authors’ claim, the visual response of F5 MNs could neither be accounted for by visually presented objects alone (Rizzolatti et al., 1988) nor for by approaching or moving stimuli of non-biological nature (Rizzolatti et al., 1981): rather, it critically required the observation of the physical interaction between an effector (e.g. the experimenter’s hand) and a manipulable object.
Figure 1 Cytoarchitectonic parcellation of the extended parieto-frontal motor system, with areas hosting neurons with mirror properties highlited in bold letters. P, principal sulcus; AI, inferior arcuate sulcus;
AS, superior arcuate sulcus; C, central sulcus; IP, intraparietal sulcus; L, lateral fissue; STS, superior temporal sulcus, Lu, Lunate sulcus.
9
Most MNs responded to the observation of only one action: thus, on the basis of the type of action activating them, they identified different pools of neurons, such as
“grasping neurons”, “manipulating neurons” or “holding neurons”. Nonetheless, the majority of MNs responded to the observations of more than one action (i.e. grasping and placing), thus demonstrating a low degree of visual specificity. The authors also stated that the response of some MNs was modulated by others action-related factors, such as action direction or the hand used by the experimenter, providing additional information concerning the observed action (Gallese et al., 1996).
One of the aspects deemed to be particularly important in the MN response was their visuo-motor congruence between the type of observed and executed action. Most MNs show some relationship between the type of action encoded visually and motorically, but according to the degree of congruence they exhibited, researchers categorized them into “strictly congruent” (when the coded action is the same in motor and visual modality), “broadly congruent” (when the visual selectivity is generally broader than the motor one) and “non-congruent” (when the coded action is different in the motor and visual modality). Because of their capacity to respond during both execution of the monkey’s own action and during the observation of a similar action performed by another individual, it was suggested that the motor system can “reflect”, like a mirror, the behavior of others’ onto one’s own behavioral repertoire. This led to propose the fascinating and now worldwide known term of “mirror neurons” for the single cells that displayed this property.
Subsequent studies demonstrated that neurons responding to others’ observed actions can also be found in the primary motor cortex (Vigneswaran et al., 2013), in the superior temporal sulcus (Perrett et al., 1989; Perrett et al., 1990; Jellema et al., 2000), in the dorsal premotor cortex (Cisek & Kalaska, 2004; Tkach et al., 2007; Dushanova &
10
Donoghue, 2010; Papadourakis & Raos, 2019), in the presupplementary motor cortex (Yoshida et al., 2011; Livi et al., 2019), in the inferior parietal lobule (Fogassi et al., 2005;
Rozzi et al., 2008), including the anterior intraparietal area (Pani et al., 2014; Maeda et al., 2015; Lanzilotto et al., 2019), in the lateral (Shepherd et al., 2009) and ventral (Ishida et al., 2010) intraparietal area, in the secondary somatosensory cortex (Hihara et al., 2015) and recently in the prefrontal cortex (Simone et al., 2015; Simone et al., 2017; Lanzilotto et al., 2017).
1.1.1 Basic properties of mirror neurons in the monkey
Many studies investigated the response properties of MNs, initially focusing demonstrating that witnessing a goal-directed action (Umiltà et al., 2001) or listening to the acoustic consequence of an invisible noisy action (Kohler et al., 2002) was necessary and sufficient to trigger MNs discharge. Particularly, Umiltà and her group recorded F5 mirror neurons while the monkey was observing a transitive action in which the critical hand-object interaction phase occurred in full vision or hidden behind an opaque screen.
In both conditions, the monkey could see whether an object was initially present or not, and hence if the experimenter’s movement was a goal-directed action or a pantomime:
importantly, the visual stimulus the monkey saw during hidden action or pantomime was exactly the same, except for the fact that the monkey was aware of whether an object was present or not.
Results showed that most of the recorded MNs responded selectively during the observation of the goal directed action relative to the pantomime even during the hidden condition, suggesting that these cells constitute the neural substrate for recognizing actions by generating an internal representation of their consequences.
11
One year later, Evelin Kohler and coworkers published a study demonstrating the existence of neurons in area F5 discharging both when the monkey performed a manual noisy action (such as peanut breaking or paper ripping) and when it heard the corresponding action-related sound in the absence of any visual information. These pioneering studies aimed to demonstrate that the essential feature of MNs was that of encoding the goal of others’ observed action, regardless of whether it could be inferred by means of a partial visual information (Umiltà et al., 2001) or by hearing the auditory feedback alone without any visual cue.
The considerable degree of generalization of the MN response across different modes of of sensory information was also supported by the evidence that the same neuron could discharge during the observation of a human hand grasping an object as well as when the grasping hand was that of a monkey (Rizzolatti et al., 1996). These results brought to the idea that each time an individual sees or listens to an action done by another individual, neurons that represent that action are automatically activated in the observer’s premotor cortex.
If observing others’ actions implies the recruitment of the same motor representations we deploy to perform similar actions, and this process involves the activation of premotor neurons, why don’t we automatically re-enact the observed action motorically? This would correspond to a known clinical sign known as echopraxia, and some neurophysiological data provide hints on what neural mechanism may prevent it from occurring during action observation. Indeed, it is known that many pyramidal tract neurons (PTNs) of the ventral premotor area F5 and of the primary motor cortex (area M1) can exhibit mirror-like properties (Kraskov et al., 2009; Vigneswaran et al., 2013;
Kraskov et al., 2014). However, while some PTNs show a classical mirror neuron response, increasing their discharge for both observation and execution of a motor act,
12
others increase their activity during action execution, but are suppressed during action observation, suggesting the existence of concurrent inhibitory mechanisms that reduce the cortico-spinal output during action observation.
The existence of such inhibitory mechanisms further suggests that neuronal repertoires of the ventral premotor cortex are decoupled from the final motor output (Bonini, 2017), at least – or in particular – during action observation, thereby preventing unwanted action execution. This allows the premotor cortex to make available its neural machinery for a variety of previously neglected generative properties including the representation of intransitive, pantomimed actions (Papadourakis & Raos, 2019), action performed with tools (Ferrari et al., 2005), physically morphed actions in which the effector (hand) is morphed into an ellipsoid moving along the same trajectory (Caggiano et al., 2016), and even withheld actions, where no movement at all is observed in a predictable phase of the trial (Bonini et al., 2014).
1.2 Evidence of a “mirror neuron system” in the human brain
The discovery of the mirror neurons in the macaque brought to the idea that such a resonance system could exist also in the human brain. Although initially there was no single-neuron study demonstrating the existence of these cells in the human brain (and no way to perform it), positron emission tomography studies showed a clear activation of the human inferior frontal gyrus (the homologue of monkey area F5) during the observation of an experimenter grasping objects (Rizzolatti et al., 1996; Grafton et al., 1996).
More evidence of the existence of a human MN system similar to the one described in monkeys came from electrophysiological studies showing a desynchronization of the mu rhythm both during the active execution of a movement and during the observation of another performing the movement (Hari et al., 1998; Cochin et
13
al., 1999). Mu-rhythm is the rhythm observed over central electrodes during EEG studies when the motor system is in a rest condition: an active movement typically suppresses it.
Ramachandran and his coworkers measured mu-rhythm in subjects observing others’
actions and equivalent movements of an inanimate object to test their hypothesis, according to which this cortical rhythm could be an indirect marker of the existence of a mirror system in humans. Indeed, they found that mu-rhythm is suppressed not only during the execution, but also during the observation of others’ movement. The mu- rhythm is therefore widely considered an electrophysiological marker of the MN activity (Altschuler et al., 1997, 2000).
Following these pioneering works, brain imaging studies provided a richer picture of the premotor areas involved in action observation. For instance, in an fMRI study (Buccino et al., 2001) subjects had to observe actions made with different effectors (mouth, hand and foot) by another individual. The authors found that observation of both object- and non-object-related actions induced a somatotopically organized activation of frontal regions rostral to the central sulcus, corresponding to the map of effectors classically attributed to the motor cortex. Authors suggested that this somatotopic activation pattern in the premotor cortex could constitute the neural substrate for a matching system between others’ action and our own motor representations. Furthermore, during the observation of object-related actions, they also found a clear activation of the inferior parietal lobule, proving that the parietal lobe is strongly activated every time an object is the target of an action.
By the way, a very consistent evidence of the existence of such a mirror mechanism in the human brain comes from TMS studies recording MEPs while stimulating the primary motor cortex of a subject observing an experimenter grasping objects: a parts of the primary motor cortex that controls particular muscles is facilitated
14
during the mere observation of actions involving these muscles (Fadiga et al., 1995;
Strafella & Paus, 2000). These results suggest that different features of the observed actions are encoded by the observer’s motor system, resulting in a facilitation of basic aspects of motor performance, such us force production (Porro et al., 2007). All these data demonstrate that the motor system allows the observer to map others’ observed actions in a motor format, thus deploying at least some of the same neural resources and representations he/she would recruit if he/she was performing that action. Indeed, humans have the ability to learn by observation alone, without concurrent practice (Mattar &
Gribble, 2005; Torriero et al., 2007), demonstrating that action observation can be a powerful tool for learning in many fields.
TMS studies also suggested that humans’ MN system can code both the goal of a motor act and the temporal aspects of every single movement constituting it: the human mirror mechanism seems to be capable of coupling action execution and action observation in terms of the detailed temporal sequence of movements to be coded (Gangitano et al., 2001).
All these studies allowed us to discover the parieto-frontal network of areas with mirror properties in the human brain (Figure 2), such as the inferior parietal lobule, the ventral premotor cortex and the inferior frontal gyrus (Caspers et al., 2010; Grosbras et al., 2012; Molenberghs et al., 2012). Furthermore, a single-subject study investigating the activation during action observation and action execution found, beside the activation of ventral premotor cortex and inferior parietal cortex, shared voxels in the supplementary motor cortex, in the middle cingulate, in the somatosensory areas, in the middle temporal cortex, in the cerebellum, in the superior parietal lobule, and in the dorsal premotor cortex, (Gazzola & Keysers, 2009), providing a more detailed description of the location of areas showing mirror properties. Particularly, these last two areas were specifically activated
15
by reaching actions, while the ventral circuit was recruited for manipulating actions, in accordance to what it is known in non-human primates.
Direct evidence of the existence of a sensory-motor mirroring mechanism in the human brain comes from a single-neuron study on implanted patients with pharmacologically intractable epilepsy (Mukamel et al., 2010), in which subjects had to observe and execute grasping actions and facial gestures during the recording of extracellular activity of several human cortical brain regions. They discovered that a significant portion of cells in the supplementary motor area and in the hippocampus discharged for both observed and executed actions, giving a clear demonstration that the same sets of neurons are activated for the production of an action and for its recognition.
On the basis of this matching between action observation and execution, and the predictive nature of these activations, the MN system is thought to be involved in the prediction of others’ impending action (Rizzolatti et al., 2014; Rizzolatti & Sinigaglia, 2010; Rizzolatti & Fogassi, 2014; Maranesi et al., 2015).
By the way, mirror response is not just a property of parieto-frontal areas involved in action representation. Such a characteristic response was also found in areas involved in behaviors with an emotional value, such as anterior (Wicker et al., 2003; Jabbi et al.,
Figure 2 Quantitative voxel-based meta-analysis showing the activation during action observation. SMA, supplementary motor area; BA6, lateral premotor cortex; BA44, 45, Broca’s area; SI, primary somatosensory area; 7A, superior parietal area; hIP3, intraparietal area; PFt, inferior parietal area;
pMTG, posterior middle temporal gyrus; V5, extrastriate visual area (Rizzolatti et al., 2014).
16
2007; Krolak-Salmon et al., 2003) and dorso-central insula (Di Cesare et al., 2014; Di Cesare et al., 2015), anterior midcingulate cortex (Hutchison et al., 1999; Singer et al., 2004), pregenual anterior cingulate cortex (Caruana et al., 2017) and amygdala. Though there is no direct evidence of the presence of this mirror mechanism in the amygdala, the existence of this “resonance” mechanism can be inferred from brain imaging and single- neuron studies. Indeed, previous studies showed an activation of this nuclear complex during the observation of fearful facial expressions (Morris et al., 1996; Phillips et al., 1997, 1998; Krolak-Salmon et al., 2003; Sato et al., 2011; Méndez-Bértolo et al., 2016) and subsequent studies demonstrated that its stimulation prompts physiological and motor responses associated with first-person fear experience (Meletti et al., 2006; Inman et al., 2018), suggesting that also amygdala neurons can show mirror properties (Livneh et al., 2012).
1.3 Evidence for view-dependent coding of observed actions
An important feature of any system devoted to recognizing observed stimuli is the capacity to process the visual features of these stimuli. Concerning others’ action, studies on the premotor cortex of the macaque investigated the influence of the viewpoint on the visual responses of mirror neurons tested with filmed actions (Caggiano et al., 2011). The findings showed that more than half of the tested MNs discharged differently for at least one point of view and a further analysis revealed that the first person was the most represented viewpoint, though the frequency of first-person selective neurons did not reach significance relative to chance. Nonetheless, when looking at LFP modulation, the authors found that the observation of an action from a subjective point of view was able to produce a significant modulation of LFP power relative to the other viewpoints, particularly in power in the low-frequency range (Caggiano et al., 2015). These findings
17
suggest that the synaptic input conveying information on the subjective viewpoint is particularly stronger in PMv.
Studies on the humans’ MN system investigated the modulation of neural response to the perspective from which an action is presented. For instance, a TMS study during the observation of intransitive movements showed that simple index and thumb movements observed from a first-person perspective elicited greater response than when observed from an allocentric perspective (Maeda et al., 2002). Furthermore, an EEG study assessing the modulation of alpha and beta mu rhythm during action observation demonstrated that also in humans the strongest sensorimotor responsiveness emerges for the subjective perspective (Angelini et al., 2018). Moreover, behavioral evidences demonstrated that the latency to imitate an observed action was significantly shorter for actions presented from a subjective perspective than from a third-person perspective (Jackson et al., 2006). Moreover, a fMRI study found a stronger and more extensive activation of the MN system when observing an action from a first-person viewpoint compared with a third-person perspective (Ge et al., 2018). By contrast, no difference in activation was obtained in a study with participants observing transitive actions (Burgess et al., 2013).
We do normally observe others’ action by a third-person perspective, so this is supposed to be the viewpoint that figures centrally in our recognition of the actions of others. Nonetheless, an fMRI study investigating this issue demonstrated that there is no evidence for spatial or attentional difference between the two perspectives (Oosterhof et al., 2012). In the cited study, subjects in the scanner were asked to look at two actions from either a first or a third-person viewpoint and then execute them. Using a multivoxel pattern analysis, authors found action-specific cross-modal visuo-motor representations in the ventral premotor cortex for the first-person perspective but not for the third-person
18
perspective, suggesting that this region is not able to generalize across both perspective and modality. On the contrary, cross-modal coding regardless of the perspective was observed in parietal and occipitotemporal area, supporting the idea that these regions have an important role in recognizing others’ observed actions.
1.4 Interaction between viewpoint and spatial location of observed action
One of the reasons why the subjective viewpoint preference of MNs initially passed unnoticed may be that the viewpoint should reasonably interact with the space sector in which a stimulus is typically observed. Indeed, it is known that the response of several MNs of area F5 can be modulated by the location in space of the observed motor act relative to the observer (Caggiano et al., 2009). Whereas part of these neurons encode space according to a metric representation, most of them encode space in operational terms, changing their properties according to the possibility for the monkey to interact with the target object and/or with the observed agent (Bonini et al., 2014). The existence of this set of mirror neurons suggest that they encode the observed motor acts to analyze those features that are relevant for generating appropriate behaviors, and the viewpoint is certainly among these features. For example, the actions we have more frequently the opportunity to observe from a subjective viewpoint in everyday life are our own actions, not those of others, and these invariably occur in the observer’s peripersonal space.
In a recent study (Maranesi et al., 2017), real observed actions were presented within the monkey’s peripersonal space from a subjective or 90° viewpoint, as well as from a 90° viewpoint far from the animal (in its extrapersonal space). Most of F5 recorded MNs exhibited a selective tuning for space during action observation, and those with peripersonal space selectivity showed a preference for the subjective rather than 90°
viewpoint. Importantly, previous studies also showed that neurons encoding other’s
19
observed actions from a subjective viewpoint are also particularly sensitive, relative to non-mirror neurons, to the visual feedback of monkey’s own hand during object grasping (Maranesi et al., 2015). These findings suggest that seeing (one’s own or another’s) hand acting into the peripersonal space may have at least two functions: it may serve as a feedback to control and monitor one’s own actions during fine tuning of behavior occurring in motor learning process (Wiggett et al., 2012; Oztop et al., 2013) and it may prompt individual’s motor response in joint actions during social interactions. Most interestingly, this effect is even stronger in MNs of the anterior intraparietal area AIP (Maeda et al., 2015; Lanzilotto et al., 2019).
1.5 Application of neuroscientific findings to human neurorehabilitation
The discovery of the “MN system” in humans and the information collected from human and non-human primate studies concerning the properties and features of MNs and the MN mechanisms, fostered the development of new neurorehabilitative approaches aimed at improving and accelerating the process of functional recovery in human patients with different types of motor impairment (Buccino et al., 2006), such as stroke (Borges et al., 2018), Parkinson’s disease (Abbruzzese et al., 2016), or cerebral palsy (Sgandurra et al., 2011), but also non-neurological patients, such as those having undergone orthopedic surgery of the hip or knee (Bellelli et al., 2010), or patients with psychiatric disorders, such as autism spectrum disorder or schizophrenia (Pineda et al., 2014; Mazza et al., 2010).
One of these new approaches deriving from MN literature is the action- observation treatment (AOT), whose aim is to restore the motor functions rather than vicariate them (Small et al., 2013): the following paragraph will trace the distinction
20
between these two types of approaches, providing examples, and finally focusing on the key aspects of AOT.
1.5.1 Rehabilitation approaches: vicariate or restoring motor functions?
The main goal of any recovery treatment after a brain damage should be to repair and promote the reorganization of the injured brain. Despite this, current practice in rehabilitation focuses mainly on the compensation of lost abilities through the re- education of the patient, for ameliorating motor or language impairments and improve the compromised functions. Most therapies come from specific theories about how certain therapeutic behaviors allow one to recover from specific impairments, hence involving focused instructions and subsequent practice that are applied to all the patients with no differences among rehabilitation style. Researchers tend to favor this more general approach, while therapists seem to prefer individually tailored instructions and practice, based on the specific needs of the patient.
Compensatory recovery seemed to be the most efficient strategy to achieve a good functional outcome (Aten et al., 1982), because it provides the tools to change the behavior to meet the new environmental needs of daily life activities by teaching the patient how to perform a task in a new way, though completely bypassing any neural restoration. Furthermore, it is a more rapid and cost-effective approach relative to restorative strategies, that require more time and huge efforts for the patient.
Several theory-driven approaches to re-education have been proposed for the treatment of different kinds of impairments, such as motor impairments or language and speech deficits. For instance, some of them have highlighted the similarity between stroke recovery and learning (Hallett, 2001; Small et al., 2002; Krakauer, 2006; Dipietro et al., 2012; Johansen-Berg et al., 2010), such as that occurring in child development. Others
21
focused on the recovery of sentence comprehension and production in patients with agrammatic Broca’s aphasia (Thompson & Shapiro, 2005) by using non-canonical sentences that implicitly train underlying properties of language that allow the patient for an effective generalization to untrained structures sharing similar linguistic properties.
It is clear that all these approaches act only on a behavioral level, affecting short- term outcomes and bypassing the main neural issue underlying the behavioral deficit.
Therefore, there was a pressing need for the development of new intervention strategies based on strong neurophysiological basis to act on the neural circuits underlying the impaired functions. Previous brain imaging data showed that both imagery and actual execution of hand actions activate similar structures of the human motor cortex (Lacourse et al., 2005; Munzert et al., 2009), suggesting that these cognitive motor processes share the same representations (Jeannerod, 2001). These findings started the development of a series of recovery strategies based on motor imagery (MI) training, an intervention approach based on the ability of humans to reproduce internally the representation of a specific action without any motor output (Decety & Grèzes, 1999). The aim of MI training (also known as mental practice) is the internal extensive reproduction of a given motor act with the intention of improving the performance (Richardson, 1967). Its positive effects have previously been assessed in several studies in sport psychology showing an improvement of the execution of movements both in athletes and novices when associated with concurred physical training (Weinberg, 1981; Suinn, 1997; Martin et al., 1999;
Cocks et al., 2014). This enhancement of the performance, although beneficial for learning new skills in novices, is more effective in elite athletes (Feltz & Landers, 1983), probably because athletes have better visualizing abilities and because they use this technique more frequently. Hence, mental practice can be a valuable tool for optimizing existing motor skills in elite athletes, as it supplements physical practice. Nonetheless,
22
mental practice alone cannot be as powerful as physical training alone, because physical training provides a series of corrective feedbacks that lead to the improvement of the performance.
Thanks to its characteristics, mental practice has been increasingly used as a complementary method for motor learning both in healthy people (Dickstein & Deutsch, 2007) and in patients with motor impairments (Malouin et al., 2013), as it seems to be a cost-efficient mean to promote motor learning and recovery. Indeed, it was demonstrated that patients with chronic effects of a stroke occurred long time before, showed a cortical reorganization after MI (Sun et al., 2013), demonstrating that this strategy acts not only at the behavioral level, but also on the neural level (Ruffino et al., 2017). Despite this, a recent systematic review (Guerra et al., 2017) and meta-analysis on stroke patients revealed the existence of a wide range of MI rehabilitation protocols and a large amount of heterogeneity in the methodological quality of the analyzed studies. Moreover, while the included studies showed positive results of the intervention based on MI, the overall meta-analysis found no significant difference compared with controls, suggesting the need of more high-quality studies and greater standardization of interventions.
As previously said, current rehabilitation approaches act on a behavioral level and then require some degree of voluntary movement. However, it is very common that people with motor impairments show a severe paresis after the brain damage and hence cannot benefit from these types of approach. An alternative approach based on visual stimulation has then been proposed: mirror therapy. This rehabilitation strategy consists in putting a mirror in the person’s midsagittal plane, thus reflecting the non-paretic side as if it were the affected one (Ramachandran et al., 1995). Movements of the non-paretic limb are able to create the illusion of the movement of the paretic limb (Deconinck et al., 2015) and its therapeutic effects have been confirmed in various disorders, including
23
stroke (Altschuler et al., 1999; Zeng et al., 2018), cerebral palsy (Park et al., 2016) and phantom pain in amputee patients (Ramachandran & Rogers, 1996).
It has been proposed that mirror therapy promotes motor function of affected limbs through the activation of the primary motor cortex (Garry et al., 2005) or of the mirror neurons (Cattaneo & Rizzolatti, 2009), though some authors found no activation of M1 or MNs in stroke patients who received MT (Michielsen et al., 2011). So, despite its efficacy is proven, the underlying mechanism is not clearly understood yet.
The following paragraph will describe one last new rehabilitation tool based on the idea that during the observation of a movement, our motor system “resonates” with the related observed action performed by another individual (Rizzolatti et al., 2001).
1.5.2 The Action Observation Treatment (AOT)
The rationale underlying the AOT is that plastic processes can occur in the motor system via the observation of meaningful daily actions performed by other healthy individuals followed by their execution - Observation to Imitate (Pomeroy et al., 2005; Garrison et al., 2010). By activating cortical motor representations during the sight of others’ bodily actions and gestures, observing others’ actions might facilitate the recovery of the compromised motor functions. This approach is based on fMRI studies showing that action observation, combined with the replication of the same action, activates the MN system stronger than when the same actions are simply observed with no prior request to imitate the observed gesture, suggesting that a learning context might facilitate the recruitment of cortical neural resources and hence boosting plasticity phenomena (Carr et al., 2003; Buccino et al., 2004).
In a typical rehabilitation session with AOT, patients watch a specific action in a video or in a live condition and then replicate what they saw in a context, practicing only
24
one action during each rehabilitation session. During the treatment, generally 20 daily actions are practiced, selected on the basis of their value in everyday life. The chosen action is normally divided into three or four motor acts and each of them is presented for 3 minutes, so that the entire duration of the video clip is almost 12 minutes. In the video, the motor acts can be performed both by an actor or an actress and they can be presented in different perspectives (subjective, frontal or lateral view, in the foreground or in the background). After the observation phase, the object used by the actor in the video clip is provided to the patients, since the observation of an object automatically recruits the most appropriate motor program to interact with it (Gibson 1977; Grafton et al., 1997; Grèzes, Tucker, et al., 2003). Patients are then required to imitate what they saw for some minutes (Buccino, 2014).
Each patient is told that the main focus of the treatment is the observation of the actions, not their execution, though they are required to perform the observed motor act to the best of their ability. Overall, an AOT rehabilitation program is supposed to last four weeks (five days a week).
This approach has been initially used with promising results in pilot studies on adults with stroke (Ertelt et al., 2007; Franceschini et al., 2010) and then widely employed in many other fields. To our knowledge, thus far, in the clinical practice, 86 studies using AOT in humans have been published (Table 1).
Neurological
disorders Orthopedic
disorders Neuropsychiatric
disorders Total
Research article 36 4 3 43
Review 12 - 1 13
Systematic review and metanalysis
5 - - 5
Randomized
controlled trial 24 1 - 25
TOTAL 77 5 4 86
Table 1
25
AOT may also be used to improve motor recovery in patients with non-neurological diseases, such as postsurgery orthopedic patients (Bellelli et al., 2010). Authors suggest that through a top-down effect, the reorganization of the motor representation at a central level occurring during AOT can influence action execution, even when the skeletal structures necessary to implement the movement are affected. Similar results were obtained in subsequent studies showing that adding action observation training to conventional therapy can increase the efficacy of rehabilitation in orthopedic patients, resulting in a greater degree of recovery (Park et al., 2014; Villafañe et al., 2017).
26 2. Objectives
The aim of this work is to analyze the efficacy of AOT on motor recovery in different clinical populations of patients. First, I will focus on the neurorehabilitation of children cerebral palsy through AOT with a meta-analytic approach. Cerebral palsy (CP) is “a group of disorders of the development of movement and posture causing activity limitation that are attributed to non-progressive disturbance that occurred in the developing fetal or infant brain. CP is often accompanied by disturbance of sensation, cognition, communication, perception, and seizure disorder” (Bax et al., 2005). It is the most common motor disability in childhood, with a prevalence of 1-3 per 1000 live births in the industrialized societies (Sellier et al., 2016). Hemiplegic forms, characterized by unilateral motor and sensory impairment, are the most frequent CPs, with a prevalence of 0.6 per 1000 live births (Johnson, 2002). Typically, the upper limb (UL) is more affected than the lower one, resulting in a significant reduction of the use of the arm and hand in daily activities, as most of them are bimanual (Sköld et al., 2004; Sakzewski et al., 2010).
Several effective interventions for the upper limb are currently available, such as intramuscular botulinum toxin A to reduce spasticity combined with UL training to improve motor skills, constraint-induced movement therapy, hand-arm bimanual intensive therapy and neurodevelopmental treatment (Sakzewski et al., 2009). But the need of new intervention strategies based on solid methodological and scientific evidences led to the development of rehabilitation protocols based on the observation of actions followed by their execution. Several randomized controlled trials that measure clinical results with AOT in children with CP have been conducted (Sgandurra et al., 2011; Buccino et al., 2012; Sgandurra et al., 2013; Kirkpatrick et al., 2016; Buccino et al., 2018; Sgandurra et al., 2018). These studies show that cerebral palsy patients benefit from AOT to improve their function.
27 3. Methods
This metanalysis on action observation training for upper limb rehabilitation in children with cerebral palsy was conducted following the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al., 2009).
Studies were searched in PubMed, with no limit for the start date as search criteria and an end date set to 13 January 2020. The keywords used to search the literature limited to human species are reported in Table 2.
One reviewer read the titles and the abstracts of the identified articles and excluded inappropriate articles on the basis of the following inclusion criteria: (1) randomized controlled trial, (2) the experimental group included action observation therapy sessions, (3) the patients were less than 18 years old, (4) the study was written in English.
The articles that could possibly fulfill the inclusion criteria were imported into Microsoft Excel and full text screening was conducted to assess whether they met the cited criteria. The following information were then extracted from each article: (1) study aim/s; (2) sample size of each group, (3) age of the participants, (4) setting, (5) duration
Recent queries in PubMed
Search Query Items found
#7 Search #5 AND #6 293
#6 Search #3 AND #4 10266
#5 Search #1 OR #2 59984
#4 Search therapy OR treatment OR training OR physical training OR rehabilitation OR neurorehabilitation OR intervention
11189643
#3 Search action observation OR action observation – execution OR
action imitation OR motor observation OR movement observation 14947
#2 Search hemipleg* OR hemipar* OR paresis OR paretic 36126
#1 Search cerebral palsy OR children cerebral palsy 26191
Table 2 Keywords used for the research in PubMed.
28
and intensity of intervention, details of the intervention received by each group, such as type of AOT and video perspective, (6) type of treatment provided to the control group, (7) outcome measures, including the mean and the standard deviation of the change in scores of outcomes before and after intervention for each group.
The data collected from the articles were analyzed using the devtools and dmetar packages with the software R, version 3.6.2 (R Core Team, 2019). Cohen d values were calculated and, according to his recommendations (Cohen, 1977), values of effect size between 0.2 and 0.5 were considered “small”, between 0.5 and 0.8 were considered
“medium” and at least 0.8 were considered “large”. In order to obtain an overall effect size, effect size estimates were pooled across the studies. Random effect models were applied to conduct the pooled analysis because the treatment durations, settings, and participants characteristics changed among the selected studies. Sidik-Jonkman estimator for 𝜏2 was calculated and Hartung-Knapp adjustment for random effects model was applied.
Cochran’s Q test and I2 percentages were used to evaluate the heterogeneity among the studies. Particularly, less than 0.05 p-values and an I2 percentage greater than 50% were interpreted as having moderate or high heterogeneity (Borenstein et al., 2009).
Finally, publication bias was assessed by funnel plot, but it was not possible to apply the Egger’s test because of the small number of studies.
29 4. Results
4.1 AOT for upper limb function in cerebral palsy 4.1.1 Identification of included studies
The database research identified 293 articles, of which 281 were excluded after the titles and the abstracts were screened. Of the 5 full papers that were read, 3 were excluded from the analysis, as they did not meet the inclusion criteria. Particularly: i) one study did not have fundamental data to conduct the metanalysis, that are means and standard deviations of the scores as a function of time in experimental and control groups and no answer to the request of these data was provided by the authors; ii) one paper had Assisting Hand Assessment (AHA) scores expressed in logit unites and could not be compared with scores reported in the other identified articles; iii) one paper was a study protocol for a subsequent study. Consequently, only two studies (Kirkpatrick et al., 2016; Buccino et al., 2018) with 86 participants could be included in this meta-analysis (Figure 3).
Figure 3 PRISMA flowchart of study selection.
30 4.1.2 Characteristics of included studies
Both selected studies focused on upper limb rehabilitation and all participants had clinical diagnosis of cerebral palsy. All participants were children and their mean age ranged from 5.6 to 7.4 years old.
In Kirkpatrick’s study AOT was conducted with home-based activities with the parent performing the required movements. In the other study (Buccino et al., 2018) AOT was undertaken with videos showing performed actions in a in-patient hospital environment. Furthermore, in the first study control group performed free actions without an observation phase, while in the second one control group watched videos showing scenes with no motor content.
The duration of AOT ranged from 3 weeks (Buccino et al., 2018) to 3 months (Kirkpatrick et al., 2016) and the amount of AOT was 15 minutes per day, 5 days a week, for both studies.
Both studies used the same outcome measure for assessing the spontaneous use of the affected hand (Assisting Hand Assessment – AHA).
4.1.3 Action observation training and control conditions
In the home-based study, a parent performed the action standing next to the child on the side of the less-affected hand facing in the same direction, so that the child observed the action from a subjective point of view, whereas in the in-patient study children observed short video clips showing an actor or an actress performing one specific daily action from different perspectives to make the video more interesting and maintain the attention of children during rehabilitation sessions.
Control group performed free actions without previous observation phase (Kirkpatrick et al., 2016), or observed videos showing images with no motor content (i.e.
31
geographical documentaries) and performed the same actions as the experimental group (Buccino et al., 2018).
4.1.4 Meta-analysis of AOT for upper limb function
A visual inspection of the funnel plot showed a nearly symmetric plot (Figure 4), suggesting absence of publication bias. By the way, it was not possible to confirm this hypothesis with the Egger’s test, because only two studies could be included in the current quantitative meta-analysis.
Figure 4 Funnel plot for included studies of the meta-analysis.
According to the random-effect model, overall effect size was not statistically significant (p = 0.14), estimated as g = 0.42 (95% CI: [-0.78; 1.61]), resulting in a moderate effect showing that AOT can be an effective tool in CP rehabilitation, though the confidence interval is very wide. The analysis of data (Figure 5) revealed that heterogeneity value was equal to zero, suggesting no relevant variability between studies. By the way, the non-significance of the p-value (p = 0.66) suggests that this result must not be taken as an evidence of no heterogeneity.
32
Figure 5 Meta-analysis of studies investigating AOT for upper limb rehabilitation in children with cerebral palsy. TE: effect size of the study; SeTE: standard error of the effect size; SMD: Hedges’g; CI:
confidence interval.
4.2 AOT in stroke patients
4.2.1 Identification of included studies
A recent systematic review (Peng et al., 2019) investigated the effectiveness of action observation treatment on several issues in patients with stroke. Authors searched the studies on several electronic databases, including PubMed, Scopus, the Cochrane Library and OTseeker (Occupational Therapy Systematic Evaluation of Evidence). They did not identify a start date for search criteria, while the end date was 31 January 2019.
Furthermore, they also used a hand search to screen whether the articles included in the previous published reviews were not included in their electronic research.
Authors stated that two reviewers read the titles and the abstract of identified articles and excluded articles that did not meet the following inclusion criteria: (1) the study was a randomized controlled trial, (2) the experimental group contained AOT sessions, (3) the stroke patients in the study were older than 18 years old, (4) the whole study was written in English or Chinese, and (5) at least one of the outcome measures was related to hand and arm function, walking ability, gait performance, or activities of daily living.
The initial database research identified 1335 articles and 3 more articles were retrieved from the references of other reviews. Finally, a total of 17 studies with 600
33
patients met all the eligibility criteria and were included in their review. By the way, 16 studies were analyzed for the quantitative meta-analysis, because the data necessary for calculating the effect sizes were not reported in one study (Figure 6).
4.2.2 Characteristics of included studies
The selected studies focused on different aspects of rehabilitation outcome, such as arm function, walking ability and daily activity performance and all participants had clinical diagnosis of stroke. The mean age of study participants ranged from 48.64 to 78.8 years old, and the stroke onset time varied from 17.8 to 1472.9 days. Moreover, the overall sample size of the included studies varied from 12 to 90 patients.
The duration of the treatment ranged from 3 to 8 weeks and the amount of AOT ranged from 20 to 90 minutes per day, three to six days a week.
Figure 6 PRISMA flowchart of study selection.
34
4.2.3 Action observation training and control conditions
Authors reported that in the selected studies AOT was conducted with videos showing arm and hand range of motion exercises, reaching and grasping movements, walking on different surfaces or in different environments, or functional, goal-directed activities.
Differently, patients in the control group received conventional rehabilitation (i.e.
occupational and/or physical therapy), sham action observation (i.e. observation of static images or pictures, such as geometric patterns or digit symbols), or they performed free actions without watching any videos.
4.2.4 Meta-analysis of AOT for arm and hand function, walking ability and activities of daily living
Authors stated that a visual inspection of the funnel plot (Figure 7) showed a nearly symmetric plot, suggesting the absence of publication bias. This hypothesis was further confirmed by the non-significant Egger’s test (t = 1.35, p = 0.20).
The analysis of the 8 articles analyzing the effect of AOT on arm and hand motor function showed a moderate significant effect size as compared with the control treatment
Figure 7 Funnel plot for all included studies of meta-analyses.
35
(Hedge’s g = 0.564; p < 0.001). Furthermore, no heterogeneity among the included studies was found (Figure 8).
Eight articles evaluated patients’ walking ability or gait performance, instead. The analysis of these data suggested that AOT has a moderate to large effect size on walking ability as compared with controls (Hedge’s g = 0.779; p < 0.001). Even in this case, no heterogeneity among the studies was found. Moreover, authors found that AOT has an average large effect on gait velocity as compared to the control treatment (Hedge’s g = 0.990; p < 0.001), and no heterogeneity among the included studies was observed (Figure 9).
Figure 8 Meta-analysis of AOT compared with the control treatment on the outcomes of arm and hand motor function
Figure 9 Meta-analysis of AOT compared with the control treatment on the outcomes of (a) walking ability and (b) gait velocity.
36
Four studies described the effects of AOT on activities of daily living. Results showed that, compared with control treatment, the AOT had a moderate to large effect size (Hedge’s g = 0.728; p = 0.004) on daily activities outcome (Figure 10). Moderate heterogeneity was found, in this case (Q = 9.258; I2 = 67.594%; df = 3).
Authors conclude that AOT can lead to moderate to large effect sizes on improving arm and hand motor function, walking ability, gait velocity and daily activity performance in both stroke patients, supporting the claim of its efficacy on the treatment of these motor impairments.
4.3 AOT in Parkinson’s disease patients
A recent systematic review (Caligiore et al., 2017) described the effects of AOT on Parkinson’s disease (PD) patients. For their analysis, authors selected two studies (Castiello et al., 2009; Pelosin et al., 2013) with different sociodemographic and clinical characteristics, as well as different experimental designs. Particularly, Castiello and coworkers used a paradigm in which both patients and healthy controls had to observe either a PD or a neurologically healthy model performing a grasping action and then had to execute the observed action.
Pelosin and her coworkers, instead, compared the effects of video and acoustic training on the execution of sequential finger movements between PD patients and
Figure 10 Meta-analysis of AOT compared with the control treatment on the outcomes of activities of daily living.
37
healthy controls and within the PD group (comparing PD patients in ON and OFF medication conditions) to analyze the degree of bradykinesia of finger movements in the different groups. Patients were randomly assigned to one of three different groups: (1) main experiment (comparing the effects of video and acoustic training), (2) control experiment 1 (observation of a video showing a static image and execution of an action), or (3) control experiment 2 (comparing the effect of video training in the ON and OFF medication states to test if the dopaminergic state of patients influenced the effect of action observation on their motor performance). Healthy subjects, instead, were recruited only for the main experiment. Authors aimed to investigate the possible different effect of AOT with respect to external rhythmic cues, so participants in the “video group” had to observe a video showing repetitive finger movements, while the participants in the
“acoustic group” had to listen for several minutes to an acoustic cue generated by a metronome and pay attention to the rhythmical cue. PD patients were tested only in the ON medication condition.
Pelosin and her coworkers found a larger increase of spontaneous movement rate after video observation than after acoustic training and a significant difference in motor performance between PD groups in ON and OFF medication states: patients not under any kind of therapy, indeed, were slower at performing finger opposition movements at baseline and after 45 minutes.
In their work, instead, Caligore and coworkers stated that in both studies PD patients undertaking AOT performed the motor tasks slower with respect to healthy controls. Specifically, the group of Castiello found a motor facilitation effect in PD patients only when the observed actor was a PD model as well. As we know, AOT exploits the same motor programs we recruit when we produce an action by ourselves;
furthermore, this activation is more intense when we observe actions belonging to our
38
motor repertoire. This can explain why patients are more facilitated in movement production when observing a pathological model: when the motor skills of the actor are more closely matched to those of the observer, there is probably an enhanced activation of the MN system. A neurophysiological evidence of this hypothesis come from a recent study (Errante et al., 2019) showing that children affected by UCP had a stronger activation of MN areas during the observation of a pathological model performing goal- directed actions.
4.4 AOT in orthopedic patients
Prior research attempted to demonstrate AOT efficacy on motor recovery in postsurgical orthopedic patients (Bellelli et al., 2010). This group applied a classical AOT paradigm in which patients were randomly assigned to case or control groups. Patients in the experimental group were asked to observe video clips showing daily actions and to imitate them afterwards; patients in the control group had to watch videos with no motor content and execute the same actions as cases. Data analysis showed a significant improvement in motor function, indicating that AOT is clearly a useful rehabilitation tool also for musculoskeletal system disorders without brain damage.
Similar results were obtained four years later by a group of researchers investigating the effect of this approach in total knee replacement (TKR) patients (Park et al., 2014). By the way, two main differences from the previous study warrant mention:
(1) the videos used in this study did not show daily life activities, but specific rehabilitation exercises to test AOT effect on gait ability, pain and activities of daily living; (2) no significant difference in gait ability was highlighted between the group receiving AOT and the physical training group.
39
It has been demonstrated that AOT can improve motor rehabilitation in TKR patients also when self-administrated (Villafañe et al., 2017). Particularly, all subjects received conventional physiotherapy and had to perform some self-administrated exercises explained in a written brochure. Patients in the experimental group had to observe a video showing a person doing some exercises, while patients in the control group had to watch a video with no motor content. The authors of this study concluded that AOT, together with conventional physiotherapy, is associated with greater degree of recovery in TKR patients.
Overall, these studies suggest that AOT can be an additional useful instrument to prompt motor recovery in orthopedic patients.
40 5. Discussion
The results of this work provide evidence for the effectiveness of AOT in improving motor function in different groups of subjects.
The main focus was to evaluate its possible efficacy on the functional recovery of upper limb in children with cerebral palsy, because no meta-analyses investigating the relevance of AOT in the rehabilitation of children with this impairment was published.
To this purpose, a meta-analysis on children with CP was performed that indicated a possible improvement of upper limb motor function following AOT with moderate effect size. This means that our findings provide some evidence that AOT may improve upper limb motor function in children with CP, but the results did not reach statistical significance. This is likely due to the small number of studies included, partly because of the restrictive inclusion criteria, but also because of the impossibility to get some fundamental data, unpublished in the related studies, which would have been necessary for including them in the quantitative meta-analysis. Another reason of non-significance may be the fact that the two included studies obtained opposite results. The group of Kirkpatrick, whose aim was to determine whether home-based parent-delivered sessions of action observation could improve upper limb motor function more than repeated practice alone, found no difference between the experimental and the control group and concluded that parent-delivered repeated practice did actually improve upper limb motor function but previous action observation is not necessary for the outcome. Buccino and coworkers, instead, found significantly stronger improvement in treated children with respect to the controls. Furthermore, they saw that this change in performance persisted also at two months follow-up, concluding that AOT is a valid tool in neurorehabilitation of children with CP.
41
A detailed analysis of the included articles showed some relevant differences between them. In Buccino’s paradigm, children in the control group had to observe videos with no motor content and then they were asked to execute the same actions as the treated participants. Differently, we noticed that in Kirkpatrick’s study the rehabilitation program of the control group did not include an observation phase (the real variable object of analysis in AOT) and, most importantly, children were not asked to perform the same actions as those in the experimental group, but they were free to play independently.
These multiple variables simultaneously manipulated makes it impossible to provide a clear vision on the real efficacy of action observation on the final outcome.
The authors of this last study stressed themselves several criticisms of their home- based rehabilitation approach: although parents in the control group were asked to encourage their children to play independently, it was not possible to strictly monitor what they actually did. Moreover, it is possible that children in the experimental group did not always observe movements before performing them, because of poor attention or lack of parental emphasis. These aspects are fundamental to obtain a reliable result that can be used for further analysis or to get information about this treatment.
This specific study did not find a significant difference between groups on the improvement of upper limb motor functions, but all the cited problems should be considered when interpreting the result. The idea behind this new approach was to provide children and families a new, less demanding, home-based rehabilitation program, but more attention should be paid to the fundamental variables that may significantly affect the final outcome.
In light of these considerations, it is then possible to better understand the non- significant result of our meta-analysis. Further studies are necessary to estimate the real
42
value of AOT in this rehabilitation field and provide more comprehensive information regarding this approach in the rehabilitation of children with CP.
The next paragraphs will describe some requirements to obtain significant improvements in motor function based on our literature research, and crucial factors to consider to design trials suitable to provide conclusive information.
5.1 Control group
As previously said, it is fundamental to have a control group to assess the real effect of a therapy on the experimental group. Subjects in the experimental group should watch short videos showing an actor performing motor acts of increasing complexity, then they should repeat what they saw. Subjects in the control group, instead, can be presented with non-action stimuli to observe (i.e. observation of landscapes, geometrical patterns, or other images with no motor content), conventional rehabilitation, or they can only practice the same movements as the case group without watching any videos. It is crucial to stress the importance of repeating the same movements patients do in the experimental group, to isolate the effect of the variable under analysis and assess if action observation has a real and specific effect on improving the performance.
5.2 Videoclips
Most of examined studies, including those that could not be included in the meta-analysis, employed videos to show patients the actions they had to perform. Videoclips are both easier to standardize and allow a broader range of patients to benefit from the therapy, like those who can perform the protocol in a home-based setting. This instrument offers a standardized model for each subject and eliminates any kind of bias of the therapist. By
